The Toxoplasma Acto-MyoA Motor Complex Is Important but Not Essential for Gliding Motility and Host Cell Invasion

Apicomplexan parasites are thought to actively invade the host cell by gliding motility. This movement is powered by the parasite's own actomyosin system, and depends on the regulated polymerisation and depolymerisation of actin to generate the force for gliding and host cell penetration. Recent studies demonstrated that Toxoplasma gondii can invade the host cell in the absence of several core components of the invasion machinery, such as the motor protein myosin A (MyoA), the microneme proteins MIC2 and AMA1 and actin, indicating the presence of alternative invasion mechanisms. Here the roles of MyoA, MLC1, GAP45 and Act1, core components of the gliding machinery, are re-dissected in detail. Although important roles of these components for gliding motility and host cell invasion are verified, mutant parasites remain invasive and do not show a block of gliding motility, suggesting that other mechanisms must be in place to enable the parasite to move and invade the host cell. A novel, hypothetical model for parasite gliding motility and invasion is presented based on osmotic forces generated in the cytosol of the parasite that are converted into motility.


Introduction
The phylum Apicomplexa consists of more than 5000 species, the majority of which are obligate intracellular parasites, and includes important human and veterinary pathogens, such as Plasmodium spp. or Toxoplasma gondii responsible for malaria and toxoplasmosis, respectively. No effective vaccines are available for use in humans and antiparasitic therapies against Apicomplexa are in general of limited efficacy due to drug resistance and lack of potency against chronic stages.
While most microbes rely on the modulation of host cell factors to trigger their uptake via endocytosis or phagocytosis, some apicomplexan parasites, such as T. gondii or Plasmodium falciparum evolved a highly complex machinery to penetrate their host cell actively within a few seconds. Although alternative mechanisms are described for other Apicomplexans, such as Theileria spp. or Cryptosposridium spp., the current theory of active host cell invasion places the actomyosin system of the parasite in the centre of the force that powers gliding motility and host cell entry [1].
The linear motor model of the parasite ( Figure 1A) predicts that the myosin A (MyoA) motor, is anchored to the inner membrane complex (IMC) located underneath the plasma membrane and moves on short actin filaments, which are connected via the glycolytic enzyme aldolase to transmembrane and surface-aligned adhesin-like proteins that act as force-transducers [2].
Smooth capping of these force-transducing molecules towards the posterior end of the parasite is associated with forward motion of the parasite on substrates or into the host cell [1]. Additional regulatory components of the MyoA motor complex include the so-called gliding associated proteins GAP40, 45 and 50 and the myosin light chain 1 (MLC1). Based on this model, it is expected that removal of these components would abolish motor function and consequently the parasite ability to locomote or invade the host cell [3][4][5]. However, video microscopy and biophysical force measurements of gliding P. berghei sporozoites suggests a more complex interplay of actin and myosin rather than a linear motor with actin controlling periodic adhesion-deadhesion cycles that account for the stick-and-slip pattern of motility [6; 7].
Host cell invasion is a multiple step process, consisting of host cell attachment, identification of the host cell by gliding motility, secretion of the rhoptries, establishment of a tight junction (TJ; a ring-like structure at the surface of the host cell) and finally active entry into the host cell through the TJ that serves as a traction point for the actomyosin system of the parasite [8].
Until recently, the traction force potential of the TJ was explained by the formation of a complex between the micronemal protein apical membrane antigen 1 (AMA1) apically secreted and the rhoptry neck protein 2 (RON2), which is a structural component of the TJ inserted into the host cell plasma membrane [9]. Furthermore AMA1 was shown to interact with actin via aldolase fulfilling therefore the requirements for acting as the specific force transmitter during host cell penetration [10]. A detailed reverse genetic analysis of parasites lacking AMA1 has however demonstrated that this protein has no critical role during force transmission, formation of the TJ or parasite entry [11]. Furthermore, preliminary results indicated that parasites lacking MyoA are viable while depletion of Act1 results in loss of the apicoplast but parasites retain some level of invasiveness. These phenotypes suggest that alternative invasion pathways that do not rely on the MyoA-actin motor may operate [12].
These findings suggest several possibilities [12]. First, multiple redundancies can exist that would complement for the removal of individual genes. In case of AMA1 this was considered less likely [11]. Similarly it is hard to imagine a functional redundancy for the single copy gene TgAct1 [12,13]. In contrast, the huge repertoire of apicomplexan myosins might show some redundancies that could complement for missing MyoA in case of myoA KO [12].
Therefore, a more complex invasion mechanism might be in place that can partially substitute for loss of a functional acto-myoA system. In this case one would expect that mutant parasites do not follow the well described step-wise process that includes formation of the typical TJ, a well accepted marker for active parasite entry [14].
To resolve this conundrum we used the DiCre regulation system [12] to engineer parasites lacking proteins of the gliding machinery that are considered as crucial to provide functional motor activity. Conditional knockouts for gap40, 45, 50, mlc1, and act1 were established and analysed in depth along with the myoA KO mutant line previously generated [12]. First we found that parasites without a functional motor complex remained motile indicating that movement can be generated in absence of the known myosin motor and parasite actin.
Secondly, none of the generated mutants showed a block in host cell invasion and in all cases entry occurred through a normally appearing TJ. Strikingly, a delay in TJ formation was detected that corresponds to the reduced overall invasion rate of the act1 KO.

Generation of conditional knockouts for mlc1, gap45 and act1
To systematically dissect the role of individual components of the gliding machinery during the asexual life cycle of the parasite, we optimised the DiCre system [12] to generate conditional knockouts (KO) in a novel DiCre Δku80 strain that shows a significantly higher efficiency of DiCre-mediated recombination upon rapamycin addition (Pieperhoff et al., in preparation). Using this strain, we generated for the first time conditional KO for gap40, gap45, gap50 and mlc1 as well as a new conditional act1 KO (Figure 1 and data not shown).
As described previously [12], these mutants were generated using the geneswap strategy, where the cDNA of the gene of interest (GOI) is flanked by loxP sides. Upon the Cremediated site-specific recombinase activity, the cDNA is removed and the reporter gene YFP placed under control of the endogenous promoter, resulting in green fluorescent KO parasites for the GOI ( Figure 1A,B). Efficiency of recombination was monitored based on the percentage of parasites expressing YFP.
All conditional KO were validated on genomic level using analytical PCR with the indicated primer pairs ( Figure 1B). Correct 5' and 3' integration was verified, and efficient Cremediated recombination was achieved in the conditional KO strains. We determined the efficiency of DiCre mediated recombination after addition of 50 nM rapamycin for 4 hours. In the case of gap45 and act1 approximately 95 % of recombination was obtained ( Figure 1B), as only the excised locus was detected in the parasite population using genomic PCRs. In the case of mlc1, the recombination rate was lower, with ~35% of the population losing the gene as judged from analytical PCR showing two bands that correspond to excised and non-excised locus, respectively ( Figure 1B).
The absence of the respective gene product in the induced population was also confirmed on protein level in immunofluorescence ( Figure 1A) or Western blot analysis ( Figure 1C). A faint signal could be observed in the induced populations for GAP45 and Act1 in western blot analysis 72 hours after removal of the respective gene. This corresponds to a minority of parasites that have not excised the floxed gene. Accordingly, in IFA analysis two distinct populations could be identified, with only very few YFP negative parasites (<10 %) that were all positive for the protein of interest. In all cases it was straight forward to identify and analyse the respective KO population, since removal of the floxed gene resulted in the activation of YFP expression leading to YFP positive KO parasites. [12] (see Figure 1A,D).
Next we performed growth assays on the conditional KO parasite lines, mlc1 KO, gap45 KO and act1 KO Therefore, parasites were induced with 50 nM rapamycin and inoculated on HFF cells. After 5 days the ability to form plaques was analysed and no growth for all three conditional KO lines was observed suggesting that MLC1, GAP45 and Act1 are essential for parasite proliferation in vitro ( Figure 1D).
Previous attempts to isolate viable KO have been successful for MIC2, MyoA and AMA1 [11,12] while they failed for mlc1, gap45 and act1, confirming the essential nature of the latter genes during the asexual life cycle. We also generated conditional KO for gap40 and gap50 (data not shown). In contrast to the other glideosome mutants, deletion of these two genes resulted in an immediate block in intracellular replication of the parasites, demonstrating that individual components of the gliding machinery can have distinct functions during the asexual life cycle. Since this study focuses on the role of the respective genes in egress, gliding and host cell invasion, data on GAP40 and GAP50 will be presented elsewhere.

Phenotypic analysis of tachyzoites lacking MyoA
It was previously suggested that MyoA is the core motor of the invasion machinery and interacts with MLC1, GAP40, GAP45 and GAP50 [3,4,15]. The phenotypic analysis of the conditional myoA KO [12], correlates to the phenotype previously described for a tetracycline-inducible KD mutant of myoA where residual host cell invasion was observed but attributed to leaky gene expression of myoA [15]. When gliding motility was analysed in a trail deposition assay, myoA KO parasites exhibited a residual gliding motility with ~ 15 % of all parasites moving by mostly circular gliding ( Figure 2A). Motility of myoA KO was further analysed using time lapse microscopy. For this 20 individual parasites were imaged and manually tracked over a time interval of 30 minutes. This kinetic analysis confirmed that myoA KO parasites moved slowly for a short distance ( Figure 2B,C) and usually stopped after ~14 µm, for several minutes, before another semicircle was formed confirming the results of the trail deposition assay (supplementary movie 1-2; Figure 2A,B,C).
Next, we compared invasion and replication rates between myoA KO and WT parasites. First, parasites were allowed to invade human foreskin fibroblasts (HFF) for different times and let to replicate for 24 hours. While no significant difference in replication was found, at each time point, invasion rate by myoA KO parasites remained constant at ~25 % compared to wt parasites ( Figure 2D), indicating that myoA KO parasites require more time to reach a similar number of invasion events as compared to RHΔhxgprt. For example the invasion rate of myoA KO parasites after 2 hours corresponded to the invasion rate of RH Δhxgprt after 30 min ( Figure 2D). Similar effects were observed using other host cells, such as HeLa cells (not shown). The overall reduction in invasion could be due to several effects. Since one of the earliest markers of the entry process is the TJ [16,17], it was important to assess if the observed reduction in invasion rate is caused by a delay in TJ formation or by a block in parasite progression into the host cell after TJ formation. While myoA KO parasites invaded the cell via a normally RON-shaped TJ ( Figure 2E), it was found in a 5 min synchronised invasion assay [18], that the majority of myoA KO parasites (~60%) remained attached to the host cell without forming a TJ. Approximately 30% of myoA KO parasites formed a TJ and another 10% of all parasites were internalized. In contrast, the majority of control parasites (80%) were found to be intracellular, 10% in the process of entry and only 10% still extracellular without initiation of TJ formation ( Figure 2F). Together these data demonstrate that a step upstream of TJ formation, such as host cell recognition or reorientation of the parasites is delayed.
To directly analyse the effect of MyoA depletion on host cell entry, the kinetics of this invasion step was analysed using time-lapse microscopy. In total 22 entry events for control and 27 for myoA KO parasites were compared ( Figure 2G). As expected, control parasites moved in within 20-30 seconds in a smooth and uniform movement (supplementary movie 3).
In contrast, myoA KO parasites showed a huge variability, with some parasites penetrating the host cell rapidly in a smooth process. However, the majority entered in a spasmodic stop-andgo fashion, and appeared stalled for several seconds to minutes (see supplementary movies 4-5; Figure 2G, H). The fastest recorded entry was ~25 seconds, whereas the slowest entry took almost 10 minutes until the parasite was completely internalised ( Figure 2H). While these results demonstrate an important function of MyoA for efficient, smooth host-cell penetration, the fact that myoA KO parasites remain capable of penetrating at similar speed as wild-type parasites indicates that the force for host-cell penetration can be generated independently of MyoA.
Together these results were interpreted that MyoA plays an important but not essential function in multiple steps during host-cell invasion. Since deletion of other components of the invasion machinery (MLC1, GAP45 and Act1) impeded mutant survival (Figure 1), we speculated that the function of MyoA might be partially complemented by other myosins.

Overlapping functions of MyoA and MyoB/C
Based on its close homology to MyoA [19], we speculated that MyoC might be capable of complementing for MyoA. Therefore, a conditional triple KO for myoA and myoB/C (note that MyoB and C are isoforms generated by alternative splicing [20] was generated by removing myoB/C in loxPMyoA ( Figure 3A). The expected genetic modifications were verified using analytical PCR on genomic DNA ( Figure 3A). No growth effect was detected for loxPMyoA-myoB/C KO parasites, indicating that myoB/C is not important for parasite survival in vitro ( Figure 3B). Induction of DiCre with 50 nM rapamycin allowed efficient removal of myoA (~60% of the population, data not shown), resulting in myoA/B/C KO parasites. Several attempts to isolate a viable triple knockout failed, indicating that removal of all three myosins is not tolerated by the parasite. Indeed, in growth assays it was observed that parasites did not form plaques in a HFF monolayer ( Figure 3B). It was found that host-cell egress is completely blocked ( Figure 3C). In contrast, albeit ~5-fold reduced when compared to myoA KO ( Figure 3D However, since it is technically not feasible to generate multiple KO for the entire repertoire of myosins that can potentially complement for MyoA, MyoB/C, we characterised other core components of the gliding and invasion machinery.

MLC1 is essential for host cell egress but not invasion
To date MLC1 is the only of the 7 myosin light chains (MLC) identified to interact with MyoA [21] and GAP40, GAP45, GAP50 in the gliding machinery and is a target for development of invasion inhibitors [5,22]. It was impossible to isolate viable mlc1 KO parasites, indicating an essential function of this gene. While MLC1 was not detectable in YFP-expressing parasites as soon as 48 hours post induction, the phenotypic analysis depicted in Figure 4 has been performed 96 h after DiCre induction to ensure absence of MLC1.
Furthermore, it was possible to maintain mlc1 KO parasites for up to 2 weeks in the induced population when artificially released from the host cell, before they were overgrown by parasites where no gene excision occurred (not shown).
We found that depletion of MLC1 had a direct effect on the localisation and expression level . We found no effect during replication after deletion of mlc1 ( Figure 4B). Next we investigated if loss of MLC1 has an impact on gliding motility. For this we only scored trails clearly deposited by YFP-expressing parasites, where mlc1 has been excised ( Figure 4C). Interestingly, mainly circular trails were observed, similar to the phenotype of myoA KO parasites. Gliding rate of mlc1 KO parasites was approximately 42 % when compared to controls ( Figure 4D). Interestingly, while parasites without MLC1 were unable to egress from the host cell when artificially induced with Ca 2+ -Ionophore ( Figure 4E), invasion rate of mlc1 KO was 28 % that of controls ( Figure 4F). As expected, mlc1 KO parasites invaded through a normally appearing TJ ( Figure 4G).
In summary mlc1 is essential for host cell egress, localisation of MyoA at the IMC, but not gliding motility and invasion. Together these data suggest that at least for MyoA function and localisation no other MLC can complement. However, it cannot be excluded that a different motor complex, such as the recently described MyoD-MLC2-motor [21] can partially complement in absence of the MyoA-motor.

GAP45 is a structural component of the IMC that is dispensable for gliding motility
GAP45 has been described as the membrane receptor for the gliding machinery [4] and its knockdown results in the relocalisation of the gliding machinery to the cytosol of the parasite in conjunction with the loss of IMC stability [3], Surprisingly some parasites remain invasive and reach relative invasion rates of 25% [3].
Since residual expression of GAP45 could not be ruled out in the KD strain, we generated an almost pure parasite population lacking GAP45 following DiCre-induced gene excision ( Figure 1A, C) and were able to keep it in culture for up to 14 days. However we did not succeed to isolate a viable gap45 KO clone. We confirmed that removal of GAP45 resulted in cytosolic localisation of other components of the gliding machinery, such as MLC1 or MyoA [3] ( Figure 5A). When compared to depletion of MyoA/B/C or MLC1, depletion of GAP45 impacted tachyzoite morphology. After exiting from their host cells, tachyzoites shifted from a crescent to a round shape, a change that was associated with the IMC detachment from the plasma membrane as observed in ultrastructural analysis ( Figure 5B). This crippled overall morphology recapitulated what has been described for gap45 KD [3] to a more pronounced extent. In contrast, intracellular parasites developed normally and showed no reduction in replication rate, as described before for gap45 KD parasites [3] ( Figure 5C).
Quantitative analysis of parasite motility revealed that gap45 KO parasites formed gliding trails similarly to control parasites ( Figure 5D,E) and that unlike previously reported, gap45 KD parasites were equally capable of gliding ( Figure 5 D,E). In addition, we analysed time lapse movies and found that depletion of GAP45 did not affect gliding speed as severely as seen in case of myoA KO since gap45 KO parasites can move double as fast (0.5 µm/s) as myoA KO parasites (0.2 µm/s)( Figure 5F).
Interestingly, it seems that morphologically less affected parasites (note: MyoA and MLC1 are cytosolic independently of parasite morphology in gap45 KO; Figure 5A) are capable to glide more efficiently, almost like control parasites. In contrast a slower, less efficient motility characterized spherical parasites (see also movies S6, 7).
Depletion of GAP45 resulted in a pronounced block of parasite egress ( Figure 5G) similarly to what we observed for mlc1 KO parasites but it caused a more severe block in host cell invasion than those seen for myoA KO or mlc1 KO (~7% when compared to control parasites, Figure 5H). Importantly, even morphologically affected, the spherical parasites were found to form a normal appearing TJ ( Figure 5I), suggesting that key features of the invasion process were preserved.
Together these data demonstrate that GAP45 has a role in providing the IMC with its typical structure thereby promoting the anchorage of the MyoA motor complex in the IMC. However, since gliding motility was less affected, it appears that the parasite can efficiently produce forward movement in the absence of a myosin motor properly anchored along the IMC.
In addition, the significant loss of invasiveness is unlikely to result from an impairment of the gliding motility as previously suggested but is rather caused by the morphologic defect of these mutants.
While the study of the motor mutants demonstrates that alternative pathways must be in place that can drive gliding motility and invasion, it does not rule out a critical function of other myosin motors. However, in case of GAP45 this motor must be sharply localised at the apical tip of the parasite which is still intact, as demonstrated by ultrastructural analysis ( Figure 5B).
In order to investigate if invasion requires a myosin motor, we investigated the role of parasite actin in more detail, since any myosin will require polymerised actin in order to exert its function as a motor.

Phenotypic analysis of act1KO parasites
We previously reported that parasite actin is essential for apicoplast replication and hence essential for the asexual life cycle of the parasite [12]. Despite that depletion of actin resulted in the loss of the apicoplast within 24 hours after removal of Act1, act1 KO parasites continued to replicate although at a slightly slower rate ( Figure 6A) before they died within vacuoles only ~ 10 days after removal of the gene, [12,24].
To analyse the involvement of actin during motility driven processes, parasites were incubated for 96 hours after activation of DiCre, which is 24 hours longer then the time required to reduce Act1 to undetectable levels as shown in immunoblot and immunofluorescence assays ( Figure 1A, 1C). It is worth mentioning that identical phenotypes were observed at earlier time points (48 hours), when residual actin was still detected (not shown), indicating that once the actin concentration is below the critical concentration, the phenotype is absolute and cannot be enhanced further. Analysis of parasite motility in trail deposition assays revealed that approximately 10% of act1 KO parasites were capable of forming trails ( Figure 6B, C and Figure S2) that in most cases corresponded to half-circles or full circles, similar to myoA KO parasites. This residual motility was completely blocked, when parasites were incubated in high potassium (endo) buffer [18] ( Figure S2). Despite maintaining a residual motility, the act1 KO tachyzoites were blocked in host cell egress after Ca 2+-Ionophore stimulation [25]; Figure 6D) and no spontaneous egress were observed over a 120 hours monitoring period ( Figure 6E). Even though in some cases parasites within the parasitophorous vacuoles started to die at this late time point, artificial release of parasites from the host cell and subsequent inoculation on fresh host cells demonstrated that parasites can re-invade new host cells. Actin was recently suggested to be involved in positioning of rhoptries to the apical complex [26]. However, when the ultrastructure of artificially released act1 KO parasites was analysed we found no morphological defects of the apical complex, in particular on rhoptry positioning, as previously reported [12]. Therefore, we conclude that the invasion defects cannot be attributed to morphological defects of the secretory organelles. It was found that the overall invasion rate was ~10 % of that control parasites ( Figure 6G). Next the ability of parasites to form a TJ was analysed as described above for myoA KO parasites ( Figure 2F). Although it has been reported that TJ formation is not blocked upon interference with parasite actin [13,27], we found that only 10% of all parasites were capable of establishing a TJ in pulse invasion assays, when compared to control parasites. This result fits well with the overall reduction in the invasion rate of act1 KO parasites. It was thus concluded that act1 KO are capable of penetrating host cells upon TJ formation ( Figure 6H) and that a step upstream of TJ formation is blocked in absence of Act1.

Discussion
Apicomplexan parasites have evolved unique organelles and structures to actively invade the host cell. The events leading to host cell invasion is a highly coordinated process that can be defined in several steps [8]: (i) parasite approach of the potential host cell by gliding (after host cell egress); (ii) host cell recognition and apical contact with the host PM; (iii); (iv) establishment of the tight junction (TJ) and finally (v) the entry into the host cell. It is widely accepted that the machinery that powers gliding motility and penetration into the host cell (v) is identical and depends on parasite actin, the MyoA motor complex and microneme proteins that act as force transmitters [28].
Conditional knock out is a powerful approach to decipher function(s) of a given protein by deleting gene in a tissue or time specific manner. First generation of these techniques has included the use of tetracycline-inducible transactivator system and has been successfully applied to Toxoplasma, in particular to the gliding machinery components [15,[29][30][31][32][33].
However, in no case, the knockdown of one of these core components resulted in the expected block in host cell invasion and this discrepancy has been explained by leaky expression of the respective gene of interest.
The recent adaptation of a conditional recombination system in T. gondii has now allowed the complete removal of components of the assumed invasion machinery, such as myoA, act1, the micronemal proteins MIC2 and AMA1 [11,12] or the rhomboid proteases ROM4 (Abstract Intriguingly, all generated mutants for the core components of the invasion machinery (see summary Table 1) remained capable of invading the host cells despite the absence of elements thought to control the function of the actin/myoA-based motor and even in absence of actin itself. While compensatory or redundancy mechanisms are likely for some of these mutants, in particular for the myosins, these results also open the possibility of alternative molecular pathways to account for gliding motility and host cell invasion.
The phenotypic analysis of myoA KO parasites suggested that MyoA plays a critical, yet not essential role for gliding motility of the parasite. Furthermore, MyoA is involved in efficient host cell entry, since many parasites enter the host cell in a slow stop-and-go fashion.
Strikingly, parasites are capable to enter at similar speed as control parasites, demonstrating that the necessary force for rapid entry can be generated in absence of MyoA. Interestingly, a triple KO for myoA,B/C resulted in a more severe phenotype compared to single myoA KO, in strong support for overlapping functions of parasite myosins. Such interchange ability and redundancy has recently been demonstrated for mouse nuclear myosin I (NM1) that can be complemented by Myo1c [34]. When the MyoA-associated and -regulatory partner MLC1 was depleted, we observed a mislocalisation of, MyoA. As expected, MLC1 loss caused a similar phenotype as observed for myoA KO parasites including a reduced gliding motility with mainly circular trails detectable and comparable invasiveness through a normally appearing TJ. In addition host cell egress was completely blocked and therefore maintenance of mlc1 KO parasites was not possible. Of note, since MLC1 depletion triggered MyoA mislocalisation, functional redundancy in the repertoire of myosin light chains is unlikely. In Strikingly, we find that gap45 KO parasites glided more efficiently than myoA KO parasites, confirming that the motility can be generated in absence of the known motor complex.
Although it is possible that a different, unknown myosin motor is involved in this process, one has to consider that the IMC, the platform for a potential second motor is disrupted in gap45 KO parasites. Finally, invasion by GAP45 depleted parasites was significantly reduced, probably as a consequence of the morphological defect but not of the loss of gliding motility.
Importantly, as described for mlc1 KO and myoA KO parasites, host cell entry proceeded through a normal TJ. Similar to mlc1 KO parasites long term cultivation of gap45 KO parasites was not possible most likely because of a block in host cell egress.
Intriguingly, depletion of parasite actin did not result in a complete block of motility, since short circular trails were readily detected in motility assays, suggesting that a residual motility is possible even in absence of parasite actin.
As is the case for motility, depletion of the core components of the known invasion machinery did not result in a block of host cell invasion. Instead it appears that the major limitation caused by depletion of this machinery lies in the delayed formation of the TJ. In case of myoA KO parasites and act1 KO parasites TJ formation was severely delayed, explaining a reduction in overall invasion rate. However, once the TJ was formed parasites entered the host cell regardless of the integrity of the typical invasion machinery. As demonstrated for myoA KO parasites, the entry process was less efficient, with many parasites moving into the host cell in a stop-and-go fashion. However, since some parasites could enter host cells at the same speed as control parasites, it is possible that the parasite or the host cell can generate the force required for entry. Together these results suggest that gliding motility is critical in a step upstream of TJ formation, but as soon as a TJ is formed the parasite can enter, albeit less efficiently, in absence of their actomyosin system. Cycles of osmotic pressure increases and relaxations at the rear could then drive protrusion.
Our observation that the gap45 KO in which effective connection of the inner membrane complex to the plasma membrane and substrate is lost, but which is still motile, argues against this mechanism.
Lastly, as is clear from our data, MyoA plays an important role in accelerating the motility and making it more steady and persistent. It is tempting to speculate that actin filament polymerize directionally (perhaps the directionality is regulated by the ion gradient at the apical tip), and then MyoA molecules could be gliding and biased to the rear of the gel, where their interactions with adhesion molecules could ratchet the cycles of gel swelling into effective protrusions of the cell leading edge. Future experiments will test whether the hypothesized osmotic engine is indeed driving gliding motility.

Cloning of DNA constructs
All primers used in this study are listed in Supplementary Table 1. To generate loxPMLC1loxP-YFP-HX the mlc1 3′UTR was amplified from genomic DNA using the primer pair 3′UTR MLC1 fw/rv, and the PCR fragment was cloned into p5RT70loxPKillerRedloxPYFP-HX [12] via SacI. The mlc1 ORF (TGME49_257680) was amplified from cDNA using the primers MLC1 ORF fw/rv, and was cloned into the parental vector using EcoRI and Pac1. To put mlc1 under the control of the endogenous promoter a 2 kb fragment upstream of the start codon of mlc1 was amplified from genomic DNA using the oligos 5′UTR MLC1 fw/rv and cloned into the parental vector using EcoRI and ApaI.
The loxPGAP45loxP-YFP-HX construct was generated using the strategy described for loxPMLC1loxP-YFP-HX with minor alterations. First, the gap45 ORF (TGME49_223940) was amplified from cDNA using the primer pair GAP45 ORF fw/rv and digested with EcoRI/PacI. Next a 2 kb fragment upstream of the start codon was amplified using

Generation and verification of parasite lines
The conditional mlc11 KO strain (ku80::diCre/endogenous mlc1::loxPmlc1loxP, referred to as loxPMLC1) was generated by transfecting 50 μg of the plasmid loxPMLC1loxPYFP-HX into the ku80::diCre [12] parasites to replace endogenous mlc1. After transfection parasites were selected for stable integration using XAN and MPA as described previously. Following the selection process the parasite pool was serially diluted to isolate single clones. The resulting loxPMLC1 strain carries only the Cre inducible copy of mlc1, allowing excision of mlc1 upon rapamycin addition (50 nM in DMSO for 4 hours prior to washout) to generate the mlc1 null mutant (ku80::diCre/mlc1 − referred to here as mlc1 KO). Confirmation of 5′ UTR integration and site-specific recombination leading to the excision of mlc1 was confirmed by PCR using the oligo set MLC1 5′UTR fw (1) and YFP rv (1'). The 3′ UTR integration into the correct locus was analysed using the primers HX fw2 (2) and MLC1 3UTR rv (2′).
To generate the conditional act1 KO and gap45 KO strains, 60 µg of the plasmids loxPAct1loxPYFP-HX and loxPGAP45loxPYFP-HX were transfected into a novel diCre ∆ku80 strain (Pieperhoff et al., in preparation) respectively. Following transfection parasites were selected for stable integration using XAN and MPA as described previously [52]. After the selection process single clones were isolated. The resulting LoxPAct1 and LoxPGAP45 strains bear only the Cre inducible copy of act11 or gap45, allowing excision of act11 and gap45 upon rapamycin addition (50 nM in DMSO for 4 hours prior to washout) to generate the act1 and gap45 KO lines (diCre ∆ku80/act11 − , diCre ∆ku80/gap45 − ) referred to here as act1 KO and gap45 KO respectively). Confirmation of 5′ UTR integration and site-specific recombination of Act1 was confirmed by PCR using the oligo set described previously [12].
To generate the conditional triple KO for myoA and myoB/C, 60 µg of the plasmid Myosin B/C KO-Bleo were transfected into the loxPMyoA strain [12] and selected for stable integration using Phleomycin as described previously [51]. Single clones were isolated resulting in the strain LoxPMyoA MyoB/C KO in which the MyoB/C locus was altered in a way that myob/c was replaced by Bleomycin. Loss of the myob/c gene was confirmed by PCR using oligo set MyoB/C gene fw/rv (3-3'). To verify the correct integration into the 5' and 3' locus the oligo pairs 5'UTR MyoB/C fw/rv (1-1') and 3'UTR MyoB/C fw/rv (2-2') were used respectively.

PCR, Western blotting and immunofluorescence analysis
For the extraction of genomic DNA from T. gondii to use as a PCR template, parasites were pelleted and afterwards washed in 1x PBS. DNA extraction was performed using Qiagen DNeasy blood and tissue kit according to manufacturer's protocol.
Western blot samples were generated by pelleting extracellular parasites and incubating them with RIPA buffer (50 mM Tris-HCl pH 8; 150 mM NaCl; 1 % Triton X-100; 0.5 % sodium deoxycholate; 0.1 % SDS; 1 mM EDTA) for 5 min on ice to lyse the parasites. Afterwards samples were centrifuged for 60 min at 14,000 rpm at 4 °C and laemmli buffer was added to the supernatant. Unless indicated otherwise 5x10 6 parasites were loaded onto a SDS acrylamide gel and immunoblot was performed as described previously [53].

Equipment and settings
For image acquisition z-stacks of 2 μm increments were collected using a UPLSAPO 100 x oil (1.40NA) objective on a Deltavision Core microscope (Image Solutions -Applied Precision, GE) attached to a CoolSNAP HQ2 CCD camera. Deconvolution was performed using SoftWoRx Suite 2.0 (Applied Precision, GE). Image acquisition was also conducted using a 100x and 63x oil objective on a Zeiss Axioskope 2 MOT+ microscope attached to an Axiocam MRm CCD camera using Volocity software, Images were processed using ImageJ 1.34r software and Photoshop (Adobe Systems Inc., USA).

Plaque assay
The plaque assay was performed as described before [15]. 200-500 freshly lysed parasites were inoculated on a confluent monolayer of HFF cells and incubated for 5 days. The HFF monolayer was washed in 1x PBS and fixed with 4 % PFA for 20 min.

Replication assay
Assay was performed as previously described [15]. In brief, 5x10 4 freshly released parasites were allowed to invade for 1 hour. Subsequently, three washing steps for removal of extracellular parasites were performed. Cells were then further incubated for 24 h before fixation. Afterwards parasites were stained with α-IMC1 antibody and the number of parasites per vacuole was counted. Mean values of three independent experiments +/ SEM were determined.

Invasion assay
In this study 2 different types of invasion assays were performed, invasion/replication or invasion/attachment. For the invasion/replication assay 5x10 4 freshly released parasites were allowed to invade for 1 hour. Subsequently, three washing steps for removal of extracellular parasites were performed. Cells were then further incubated for 24 h before fixation. Mean values of three independent assays +/ SEM were determined.

Gliding assay
Trail deposition assays have been performed as described before [54]. Briefly, freshly released parasites were allowed to glide on FBS-coated glass slides for 30 min before they were fixed with 4% PFA and stained with SAG1.

Time lapse microscopy
Time-lapse video microscopy was conducted with the DeltaVision ® Core microscope using a 40X objective lens for invasion analysis and 20X objective for gliding analysis. Normal growth conditions were maintained throughout the experiment (37 °C; 5 % CO 2 ). Images were recorded at one frame per second using SoftWoRx® software. Further image processing was performed using ImageJ 1.34r software.
In order to assess penetration time, heavily infected HFFs were scratched and the cells suspension was passed through 26 gauge needle three times to release parasites artificially. About 1 × 10 6 parasites were then added onto HFFs grown in Ibidi µ-Dish 35mm, high . Invasion events were observed after about 20 min when parasites had settled and corresponding penetration time was determined. To assess the gliding kinetics parasites were prepared akin to the trail deposition assay. Briefly, extracellular parasites were pelleted, washed in prewarmed PBS and resuspended to a concentration of 1 × 10 6 per 800 µl in pre-warmed HBSS.
Parasites were then added to FCS-coated glass dishes (Ibidi µ-Dish 35mm, high ) and gliding was recorded for 30 min after parasites have settled. For analysis, 20 parasites were manually tracked using the manual tracking plugin for ImageJ 1.34r software and total displacement as well as average speed were determined.

Staining and quantification of tight junction
Rhoptry secretion and tight junction formation was assessed as described by [55] with minor changes Briefly, 36 hours post inoculation heavily infected HFFs were scratched and parasites were released by passing the cell suspension through a 26 gauge needle followed by centrifugation at 600 g for 5 min. Supernatant was removed and parasites were resuspended in 3 × 10 6 /200 µl Endo buffer (44.7 mM K 2 SO 4 , 10 mM Mg 2 SO 4 , 106 mM sucrose, 5 mM glucose, 20 mM Tris, 0.3 5% (w/v) BSA, pH 8.2). Parasites were incubated at RT for 10 min before adding 200 µl to a confluent monolayer of HFFs grown on coverslips. Parasites and cells were incubated for an additional 3 min at RT followed by 20 min at 37 °C and 5 % CO 2 to let the parasite settle naturally. Please note that parasites are unable to secrete their micronemes in high potassium buffer, thus attachment is only lose. Therefore, supernatant was carefully replaced by 500 µl pre-warmed growth media and parasites were let to invade for 5 min at normal growth conditions. After three washing steps with PBS, the cells were then fixed with 4 % paraformaldehyde followed by immunostaining under non-permeabilising conditions with rabbit α-Toxoplasma and mouse α-RON4 primary antibody followed by Alexa Fluor secondary antibody, respectively. Alexa Fluor 488/594 secondary antibody was used at 1:500 dilution to increase the RON4 signal at the tight junction. Assay was performed in triplicate at three independent occasions and 200 parasites were counted and calculated as percentage.

Electron Microscopy
Samples of extra-and intra-cellular tachyzoites of wild type and GAP45 KO cultured for 48 and 72 hours were fixed with 2.5 % gluteraldehyde in 0.1 M phosphate buffer pH 7.4 (1M Na2HPO4, 1M NaH2PO4). Samples were then processed for routine electron microscopy as